The Vibe Coding Revolution Is Cancelled

0. Introduction

I refuse to compare the pros and cons of vibe coding with traditional programming, because "nothing has happened that would significantly change the situation." This article develops an idea from my previous work: that reality changes faster than our understanding of those changes.

1. Definitions

The term "vibe coding" was coined by OpenAI co-founder Andrei Karpathy in March 2025. Karpathy described it as programming where the developer "completely surrenders to the vibes" and doesn't think about the code. Key characteristics:

• "I just chat with Composer using SuperWhisper, so I barely touch the keyboard"
• The developer asks the computer to perform simple tasks ("reduce the sidebar padding by half")
• Doesn't read the proposed changes (diffs)
• Copies error messages and pastes them into the chat without comment
• "I'm building a project or web app, but it's not really programming"

2. Vibe Coders of Antiquity

I argue that similar processes already existed decades ago.

HTML and Web Development: There was a time when deep knowledge of HTML tags was required. But then came:

• Visual editors (Dreamweaver)
• WYSIWYG editors in CMS platforms
• Drag-and-drop design interfaces
• The ability to create HTML in rich-text editors

All of this fits the description of vibe coding: the user doesn't think about the code, gives commands, things constantly "crawl and bulge," but the author manages to deal with errors step by step.

Analogy with office suites: Editing complex text in Word is also vibe coding — the result never matches expectations.

Objection: "HTML coders using Dreamweaver... that's not the same as real coders who bear the proud title of 'programmer.'" But this distinction was relevant "until February 3, 2025, when the accursed OpenAI gave all idiots the ability to code."

3. Real Programmers of Today

Let's analyze what ordinary programmers actually do:

Using built-in functions: Modern developers work with high-level languages featuring built-in functions whose internals they don't understand. They say "sort the array" without knowing which algorithm is used, and make mistakes when using these functions.

Syntax: Programmers don't memorize syntax — they rely on IDEs with autocompletion and hints. Essentially, they're asking the computer to insert a function template.

Libraries and frameworks: They use third-party components written by other people they've never met. The code boils down to library calls and a single launch command.

Finding solutions: Programmers turn to documentation, other people's code, or Stack Overflow, copy suitable code, run it, look at the errors, and tweak it.

Conclusion: The only difference from vibe coding is voice control using natural language. The process of "copying and tweaking" has existed for a very long time. Using natural language is simply a separate skill, similar to knowing how to search documentation or how to google effectively.

4. So What Is "Vibe Coding" Really?

My blunt answer: it's a marketing term that OpenAI uses to achieve higher status in the industry. When the term is applied to other companies' tools, someone will inevitably pop up claiming that "vibe coding was invented by OpenAI."

What actually changed: The process of change is continuous, but the processes associated with vibe coding were already present in "traditional programming" and are now simply "even more mechanized and automated."

The myth of unqualified coder invasion: There's no need to fear that LLM tools will open the industry to incompetent developers producing cheap, low-quality code. Such people have long been in the industry and have long been producing low-quality code. Their appearance is due not to new tools, but to the "growing need for workers of this profile."

The true role of LLMs: Large language model assistants are "simply an answer to the demand 'we need more coders.'" They are not a "driver of significant changes," but rather their "consequence."

Final verdict: The revolution is cancelled.